Completed Thesis Projects

The battle of developing hardware Trojans and corresponding countermeasures has taken adversaries towards ingenious ways of compromising hardware designs by circumventing even advanced testing and verification methods. Besides conventional methods of inserting Trojans into a design by a malicious entity, the design flow for field-programmable gate arrays (FPGAs) can also be surreptitiously compromised to assist the attacker to perform a successful malfunctioning or information leakage attack. This thesis mainly focuses on the two aspects of hardware Trojans in reconfigurable systems, the defenders perspective which corresponds to the bitstream-level Trojan detection technique, and the attackers perspective which corresponds to a novel FPGA Trojan attack. From the defender's perspective, we introduce a first-ever successful pre-configuration countermeasure against the ``Malicious LUT''-hardware Trojan, by employing bitstream-level Proof-Carrying Hardware (PCH) and present the complete design-and-verification flow for iCE40 FPGAs. Likewise, from an attackers perspective, we present a novel attack that leverages malicious routing of the inserted Trojan circuit to acquire a dormant state even in the generated and transmitted bitstream. Since the Trojan is injected in a post-synthesis step and remains unconnected in the bitstream, the presented attack can currently neither be prevented by conventional testing and verification methods nor by bitstream-level verification techniques.

Previous research in proof-carrying hardware has established the feasibility and utility of the approach, and provided a concrete solution for employing it for the certification of functional equivalence checking against a specification, but fell short in connecting it to state-of-the-art formal verification insights, methods and tools. Due to the immense complexity of modern circuits, and verification challenges such as the state explosion problem for sequential circuits, this restriction of readily-available verification solutions severely limited the applicability of the approach in wider contexts. This thesis closes the gap between the PCH approach and current advances in formal hardware verification, provides methods and tools to express and certify a wide range of circuit properties, both functional and non-functional, and presents for the first time prototypes in which circuits that are implemented on actual reconfigurable hardware are verified with PCH methods. Using these results, designers can now apply PCH to establish trust in more complex circuits, by using more diverse properties which they can express using modern, efficient property specification techniques.

Traditional cache design uses a consolidated block of memory address bits to index a cache set, equivalent to the use of modulo functions. While this module-based mapping scheme is widely used in contemporary cache structures due to the simplicity of its hardware design and its good performance for sequences of consecutive addresses, its use may not be satisfactory for a variety of application domains having different characteristics.This thesis presents a new type of cache mapping scheme, motivated by programmable capabilities combined with Nature-inspired optimization of reconfigurable hardware. This research has focussed on an FPGA-based evolvable cache structure of the first level cache in a multi-core processor architecture, able to dynamically change cache indexing. To solve the challenge of reconfigurable cache mappings, a programmable Boolean circuit based on a combination of Look-up Table (LUT) memory elements is proposed. Focusing on optimization aspects at the system level, a Performance Measurement Infrastructure is introduced that is able to monitor the underlying microarchitectural metrics, and an adaptive evaluation strategy is presented that leverages on Evolutionary Algorithms, that is not only capable of evolving application-specific address-to-cache-index mappings for level one split caches but also of reducing optimization times. Putting this all together and prototyping in an FPGA for a LEON3/Linux-based multi-core processor, the creation of a system architecture reduces cache misses and improves performance over the use of conventional caches.

The use of heterogeneous computing resources, such as graphics processing units or other specialized co-processors, has become widespread in recent years because of their performance and energy efficiency advantages. Operating system approaches that are limited to optimizing CPU usage are no longer sufficient for the efficient utilization of systems that comprise diverse resource types. Enabling task preemption on these architectures and migration of tasks between different resource types at run-time is not only key to improving the performance and energy consumption but also to enabling automatic scheduling methods for heterogeneous compute nodes. This thesis proposes novel techniques for run-time management of heterogeneous resources and enabling tasks to migrate between diverse hardware. It provides fundamental work towards future operating systems by discussing implications, limitations, and chances of the heterogeneity and introducing solutions for energy- and performance-efficient run-time systems. Scheduling methods to utilize heterogeneous systems by the use of a centralized scheduler are presented that show benefits over existing approaches in varying case studies.

Monte-Carlo Tree Search (MCTS) is a class of simulation-based search algorithms. It brought about great success in the past few years regarding the evaluation of deterministic two-player games such as the Asian board game Go. In this thesis, we present a parallelization of the most popular MCTS variant for large HPC compute clusters that efficiently shares a single game tree representation in a distributed memory environment and scales up to 128 compute nodes and 2048 cores. It is hereby one of the most powerful MCTS parallelizations to date. In order to measure the impact of our parallelization on the search quality and remain comparable to the most advanced MCTS implementations to date, we implemented it in a state-of-the-art Go engine Gomorra, making it competitive with the strongest Go programs in the world. We further present an empirical comparison of different Bayesian ranking systems when being used for predicting expert moves for the game of Go and introduce a novel technique for automated detection and analysis of evaluation uncertainties that show up during MCTS searches.

Handling run-time dynamics on embedded system-on-chip architectures has become more challenging over the years. On the one hand, the impact of workload and physical dynamics on the system behavior has dramatically increased. On the other hand, embedded architectures have become more complex as they have evolved from single-processor systems over multi-processor systems to hybrid multi-core platforms.Static design-time techniques no longer provide suitable solutions to deal with the run-time dynamics of today's embedded systems. Therefore, system designers have to apply run-time solutions, which have hardly been investigated for hybrid multi-core platforms.In this thesis, we present fundamental work in the new area of run-time management on hybrid multi-core platforms. We propose a novel architecture, a self-adaptive hybrid multi-core system, that combines heterogeneous processors, reconfigurable hardware cores, and monitoring cores on a single chip. Using self-adaptation on thread-level, our hybrid multi-core systems can effectively perform performance and thermal management autonomously at run-time.

Reconfigurable circuit devices have opened up a fundamentally new way of creating adaptable systems. Combined with artificial evolution, reconfigurable circuits allow an elegant adaptation approach to compensating for changes in the distribution of input data, computational resource errors, and variations in resource requirements. Referred to as ``Evolvable Hardware'' (EHW), this paradigm has yielded astonishing results for traditional engineering challenges and has discovered intriguing design principles, which have not yet been seen in conventional engineering. In this thesis, we present new and fundamental work on Evolvable Hardware motivated by the insight that Evolvable Hardware needs to compensate for events with different change rates. To solve the challenge of different adaptation speeds, we propose a unified adaptation approach based on multi-objective evolution, evolving and propagating candidate solutions that are diverse in objectives that may experience radical changes. Focusing on algorithmic aspects, we enable Cartesian Genetic Programming (CGP) model, which we are using to encode Boolean circuits, for multi-objective optimization by introducing a meaningful recombination operator. We improve the scalability of CGP by objectives scaling, periodization of local- and global-search algorithms, and the automatic acquisition and reuse of subfunctions using age- and cone-based techniques. We validate our methods on the applications of adaptation of hardware classifiers to resource changes, recognition of muscular signals for prosthesis control and optimization of processor caches.

FPGAs, systems on chip and embedded systems are nowadays irreplaceable. They combine the computational power of application specific hardware with software-like flexibility. At runtime, they can adjust their functionality by downloading new hardware modules and integrating their functionality. Due to their growing capabilities, the demands made to reconfigurable hardware grow. Their deployment in increasingly security critical scenarios requires new ways of enforcing security since a failure in security has severe consequences. Aside from financial losses, a loss of human life and risks to national security are possible. With this work I present the novel and groundbreaking concept of proof-carrying hardware. It is a method for the verification of properties of hardware modules to guarantee security for a target platform at runtime. The producer of a hardware module delivers based on the consumer's safety policy a safety proof in combination with the reconfiguration bitstream. The extensive computation of a proof is a contrast to the comparatively undemanding checking of the proof. I present a prototype based on open-source tools and an abstract FPGA architecture and bitstream format. The proof of the usability of proof-carrying hardware provides the evaluation of the prototype with the exemplary application of securing combinational and bounded sequential equivalence of reference monitor modules for memory safety.

The paradigm shift towards many-core parallelism is accompanied by two fundamental questions: how should the many processors on a single die communicate to each other and what are suitable programming models for these novel architectures? In this thesis, the author tackles both questions by reviewing the reconfigurable mesh model of massively parallel computation for many-cores. The book presents the design, implementation and evaluation of a many-core architecture that is based on the execution principles and communication infrastructure of the reconfigurable mesh. This work fundamentally rests on FPGA implementations and shows that reconfigurable mesh processors with hundreds of autonomous cores are feasible. Several case studies demonstrate the effectiveness of programming and illustrate why the reconfigurable mesh is a promising model for many-cores.